Research Of Digital Image Processing Parallization Based On Multicore DSP

Digital Signal Processer (DSP) is one of those microprocessor specifically designedfor realizing digital signal processing. DSPs are widely used in digital image processingsystem, especially in those biomedical engineering system such as medical image systemlike CT, MRI and USG etc. or high definition medical image compressing and deliverysystem or medical image reconstruction system.The multicore DSPs bring the performance of DSP’s to a new stage. But there are stilla lot of problem in programming a parallel executed program which can really takeadvantage of the multicore processer. The conventional parallel programing methods arebased on those general parallel computing system and are not optimized for the hardwarearchitecture of DSP’s. This paper did a research on the conventional parallel programingmethods and designed a parallelization method for the digital image processing whichexecute on a multicore DSP.Firstly, an introduction is given about the development of multicore processer andparallel programming. Then, detailed introductions are given about the architecture ofmulticore processer and parallel programming models. Descriptions of parallel programingmethods as long as the evaluation standards and the implementation procedures are alsoincluded. A digital image processing platform was built based on the multicore DSPTMS320C6678which is briefing in the paper at the first place. Problems that will appear inthe implementation of parallel algorism on this platform are analyzed in detail and relevantsolutions are given. A digital image processing parallelization method which is optimizedfor the multicore DSP is put forward. Finally, an image matching algorithm based oncorrelation is implemented on the image processing platform using the newly proposedparallelization method. The result shows that the parallelization method works correctly onthe multicore image processing platform and the execution speed are increased effectively.